Target detection method based on multi-source feature joint network and transformed image self-generation

Abstract

The invention discloses a target detection method based on a multi-source feature joint network and transformation image self-generation. The method comprises the following steps: respectively utilizing an initial target detection network for a to-be-detected infrared image and a to-be-detected visible light image in the same scene to obtain initial target detection results; obtaining a multi-source feature correction parameter of the to-be-detected infrared image relative to the to-be-detected visible light image by using the multi-source feature joint network; calculating position coordinates of each target coordinate relative to the visible light image to be detected in the initial target detection result of the infrared image to be detected by using the multi-source feature correction parameters to obtain a first set of infrared relative visible light target coordinates; for the first set and the second set, judging whether every two elements in the two sets belong to the same target or not by calculating the intersection-to-union ratio of the two elements, and obtaining a final target detection result according to a judgment result; the target detection time can be shortened, the data set generation efficiency is improved, the cost is reduced, and the accuracy is improved.

Description

technical field [0001] The invention belongs to the technical field of target detection, and in particular relates to a target detection method based on multi-source feature joint network and transformed image self-generation. Background technique [0002] In order to overcome the problem that the observation performance of a single sensor is severely limited by severe weather conditions, multi-source means can usually be used to achieve complementary advantages. In recent years, many researchers have achieved certain results in the field of target detection and fusion of multi-source sensors. The representative way is to perform pixel registration on multi-source sensor data, and then perform fusion at the data level, and then obtain Multi-source sensor fusion image for object detection. [0003] However, this traditional method cannot meet the needs of real-time detection and fusion due to the complicated algorithm steps of detection and fusion and high computational cost...

AI Technical Summary

Problems solved by technology

[0003] However, this traditional method cannot meet the needs of real-time detection and fusion due to the complicated algorithm steps of detection and fusion and high computational cost, so it is not suitable for scenarios with high real-time requirements such as search and rescue of disaster personnel

Images